Adolescent Non-Suicidal Self-Injury Risk Factors for Machine Learning Models

A retrospective cohort of 588 adolescent patients who received psychological and psychiatric assessments. The data includes candidate predictors such as demographic characteristics, psychological and emotional status, behavioral characteristics, and peer support, used to develop a machine learning risk prediction model. The dataset was created by Yujun Zhao and last updated on 2026-05-13.

Use Cases

• Training a support vector machine model to predict non-suicidal self-injury risk based on psychological and behavioral factors.
• Comparing the performance of multiple machine learning models (MLR, RF, SVM, XGBoost) for clinical prediction tasks.
• Conducting interpretability analysis using SHAP to identify key predictors like suicide-related ideation and school bullying.
• Developing early identification tools for adolescents at high risk of self-injury based on demographic and emotional status data.

Strengths

• Includes 588 patient records, providing a basis for model training and validation.
• Model performance was evaluated at three time points (T1, T2, T3) with reported incidence rates of approximately 24%, 23%, and 22%.
• The optimal SVM model demonstrated AUC values all greater than 0.75 and recall and F1 scores both higher than 0.7.
• Data covers multiple predictor categories including demographics, psychological status, behavior, and peer support.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is very small at 19.7 KB, indicating limited scope.

Provenance

Source

Yujun Zhao

Collection Method

Retrospective cohort study design using psychological and psychiatric assessments.

Freshness

Last updated 2026-05-13 05:51:47; freshness should be verified.